Skip to content

Return missing when calculating area_under_curve() for empty inputs #49

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hannahilea merged 7 commits into from
Mar 15, 2022
Merged

Return missing when calculating area_under_curve() for empty inputs #49

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
27b2b8a
throw don't assert
hannahilea Mar 15, 2022
71f4a76
Remove additional assert
hannahilea Mar 15, 2022
3205eeb
Fix #16
hannahilea Mar 15, 2022
f9eb37b
Bump patch version
hannahilea Mar 15, 2022
30cd2f5
fix style
hannahilea Mar 15, 2022
c325735
whoops
hannahilea Mar 15, 2022
9cac3b3
fix style
hannahilea Mar 15, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
name = "Lighthouse"
uuid = "ac2c24cd-07f0-4848-96b2-1b82c3ea0e59"
authors = ["Beacon Biosignals, Inc."]
version = "0.13.3"
version = "0.13.4"

[deps]
Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
Expand Down
2 changes: 1 addition & 1 deletion src/metrics.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,7 +105,7 @@ Return `(κ, p₀)` where `κ` is Cohen's kappa and `p₀` percent agreement giv
their equivalents in [`confusion_matrix`](@ref)).
"""
function cohens_kappa(class_count, hard_label_pairs)
@assert all(issubset(pair, 1:class_count) for pair in hard_label_pairs)
all(issubset(pair, 1:class_count) for pair in hard_label_pairs) || throw(ArgumentError("Unexpected class in `hard_label_pairs`."))
p₀ = accuracy(confusion_matrix(class_count, hard_label_pairs))
pₑ = _probability_of_chance_agreement(class_count, hard_label_pairs)
return _cohens_kappa(p₀, pₑ), p₀
Expand Down
11 changes: 6 additions & 5 deletions src/utilities.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,17 +13,18 @@ end
area_under_curve(x, y)

Calculates the area under the curve specified by the `x` vector and `y` vector
using the trapezoidal rule.
using the trapezoidal rule. If inputs are empty, return `missing`.
"""
function area_under_curve(x, y)
@assert length(x) == length(y)
length(x) == length(y) || throw(ArgumentError("Length of inputs must match."))
length(x) == 0 && return missing
auc = zero(middle(one(eltype(x)), one(eltype(y))))
perms = sortperm(x)
sorted_x = view(x, perms)
sorted_y = view(y, perms)
# calculate the trapazoidal method https://en.wikipedia.org/wiki/Trapezoidal_rule
for i in 2:length(x)
auc += middle(sorted_y[i], sorted_y[i - 1]) * (sorted_x[i] - sorted_x[i - 1])
auc += middle(sorted_y[i], sorted_y[i-1]) * (sorted_x[i] - sorted_x[i-1])
end
return auc
end
Expand All @@ -32,10 +33,10 @@ end
area_under_curve_unit_square(x, y)

Calculates the area under the curve specified by the `x` vector and `y` vector
for a unit square, using the trapezoidal rule.
for a unit square, using the trapezoidal rule. If inputs are empty, return `missing`.
"""
function area_under_curve_unit_square(x, y)
@assert length(x) == length(y)
length(x) == length(y) || throw(ArgumentError("Length of inputs must match."))
kept = [(i, j)
for (i, j) in zip(x, y)
if !(ismissing(i) || ismissing(j)) && (0 <= i <= 1 && 0 <= j <= 1)]
Expand Down
62 changes: 32 additions & 30 deletions test/metrics.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,22 +49,22 @@
c = confusion_matrix(k, hard_label_pairs)
kappa, percent_agreement = cohens_kappa(3, hard_label_pairs)
@test percent_agreement == accuracy(c)
@test isapprox(percent_agreement, 0.5; atol=0.02)
@test isapprox(kappa, 0.0; atol=0.02)
@test isapprox(percent_agreement, 0.5; atol = 0.02)
@test isapprox(kappa, 0.0; atol = 0.02)
stats = binary_statistics(c, 1)
@test isapprox(stats.predicted_positives, 500_000; atol=2000)
@test isapprox(stats.predicted_negatives, 500_000; atol=2000)
@test isapprox(stats.actual_positives, 500_000; atol=2000)
@test isapprox(stats.actual_negatives, 500_000; atol=2000)
@test isapprox(stats.true_positives, 250_000; atol=2000)
@test isapprox(stats.true_negatives, 250_000; atol=2000)
@test isapprox(stats.false_positives, 250_000; atol=2000)
@test isapprox(stats.false_negatives, 250_000; atol=2000)
@test isapprox(stats.true_positive_rate, 0.5; atol=0.02)
@test isapprox(stats.true_negative_rate, 0.5; atol=0.02)
@test isapprox(stats.false_positive_rate, 0.5; atol=0.02)
@test isapprox(stats.false_negative_rate, 0.5; atol=0.02)
@test isapprox(stats.precision, 0.5; atol=0.02)
@test isapprox(stats.predicted_positives, 500_000; atol = 2000)
@test isapprox(stats.predicted_negatives, 500_000; atol = 2000)
@test isapprox(stats.actual_positives, 500_000; atol = 2000)
@test isapprox(stats.actual_negatives, 500_000; atol = 2000)
@test isapprox(stats.true_positives, 250_000; atol = 2000)
@test isapprox(stats.true_negatives, 250_000; atol = 2000)
@test isapprox(stats.false_positives, 250_000; atol = 2000)
@test isapprox(stats.false_negatives, 250_000; atol = 2000)
@test isapprox(stats.true_positive_rate, 0.5; atol = 0.02)
@test isapprox(stats.true_negative_rate, 0.5; atol = 0.02)
@test isapprox(stats.false_positive_rate, 0.5; atol = 0.02)
@test isapprox(stats.false_negative_rate, 0.5; atol = 0.02)
@test isapprox(stats.precision, 0.5; atol = 0.02)

@test confusion_matrix(10, ()) == zeros(10, 10)
@test all(ismissing, cohens_kappa(10, ()))
Expand All @@ -91,6 +91,8 @@
@test Lighthouse._cohens_kappa(p, p / 2) > 0
end
end

@test_throws ArgumentError cohens_kappa(3, [(4,5),(8,2)])
end

@testset "`calibration_curve`" begin
Expand All @@ -99,42 +101,42 @@ end
bitmask = rand(rng, Bool, 1_000_000)
bin_count = 12
bins, fractions, totals, mean_squared_error = calibration_curve(probs, bitmask;
bin_count=bin_count)
bin_count = bin_count)
@test bin_count == length(bins)
@test first(first(bins)) == 0.0 && last(last(bins)) == 1.0
@test all(!ismissing, fractions)
@test all(!iszero, totals)
@test all(isapprox.(fractions, 0.5; atol=0.02))
@test all(isapprox.(totals, length(probs) / bin_count; atol=1000))
@test all(isapprox.(fractions, 0.5; atol = 0.02))
@test all(isapprox.(totals, length(probs) / bin_count; atol = 1000))
@test sum(totals) == length(probs)
@test isapprox(ceil(mean(fractions) * length(bitmask)), count(bitmask); atol=1)
@test isapprox(mean_squared_error, inv(bin_count); atol=0.002)
@test isapprox(ceil(mean(fractions) * length(bitmask)), count(bitmask); atol = 1)
@test isapprox(mean_squared_error, inv(bin_count); atol = 0.002)

rng = StableRNG(42)
probs = range(0.0, 1.0; length=1_000_000)
probs = range(0.0, 1.0; length = 1_000_000)
bitmask = [rand(rng) <= p for p in probs]
bin_count = 10
bins, fractions, totals, mean_squared_error = calibration_curve(probs, bitmask;
bin_count=bin_count)
ideal = range(mean(first(bins)), mean(last(bins)); length=bin_count)
bin_count = bin_count)
ideal = range(mean(first(bins)), mean(last(bins)); length = bin_count)
@test bin_count == length(bins)
@test first(first(bins)) == 0.0 && last(last(bins)) == 1.0
@test all(!ismissing, fractions)
@test all(!iszero, totals)
@test all(isapprox.(fractions, ideal; atol=0.01))
@test all(isapprox.(fractions, ideal; atol = 0.01))
@test all(totals .== 1_000_000 / bin_count)
@test isapprox(ceil(mean(fractions) * length(bitmask)), count(bitmask); atol=1)
@test isapprox(mean_squared_error, 0.0; atol=0.00001)
@test isapprox(ceil(mean(fractions) * length(bitmask)), count(bitmask); atol = 1)
@test isapprox(mean_squared_error, 0.0; atol = 0.00001)

bitmask = reverse(bitmask)
bins, fractions, totals, mean_squared_error = calibration_curve(probs, bitmask;
bin_count=bin_count)
bin_count = bin_count)
@test bin_count == length(bins)
@test first(first(bins)) == 0.0 && last(last(bins)) == 1.0
@test all(!ismissing, fractions)
@test all(!iszero, totals)
@test all(isapprox.(fractions, reverse(ideal); atol=0.01))
@test all(isapprox.(fractions, reverse(ideal); atol = 0.01))
@test all(totals .== 1_000_000 / bin_count)
@test isapprox(ceil(mean(fractions) * length(bitmask)), count(bitmask); atol=1)
@test isapprox(mean_squared_error, 1 / 3; atol=0.01)
@test isapprox(ceil(mean(fractions) * length(bitmask)), count(bitmask); atol = 1)
@test isapprox(mean_squared_error, 1 / 3; atol = 0.01)
end
20 changes: 11 additions & 9 deletions test/utilities.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,23 +3,25 @@
@test majority([1, 2, 1, 3, 2, 2, 3, 4], 3:4) == 3
rng = StableRNG(42)
picked = [majority(rng, 1:2, 1:2) for _ in 1:1_000_000]
@test isapprox(count(==(1), picked), 500_000; atol=1000)
@test isapprox(count(==(1), picked), 500_000; atol = 1000)
end

@testset "`Lighthouse.area_under_curve`" begin
@test_throws AssertionError Lighthouse.area_under_curve([0, 1, 2], [0, 1])
@test_throws ArgumentError Lighthouse.area_under_curve([0, 1, 2], [0, 1])
@test ismissing(Lighthouse.area_under_curve([], []))
@test isapprox(Lighthouse.area_under_curve(collect(0:0.01:1), collect(0:0.01:1)), 0.5;
atol=0.01)
atol = 0.01)
@test isapprox(Lighthouse.area_under_curve(collect(0:0.01:(2π)), sin.(0:0.01:(2π))),
0.0; atol=0.01)
0.0; atol = 0.01)
end

@testset "`Lighthouse.area_under_curve_unit_square`" begin
@test_throws AssertionError Lighthouse.area_under_curve_unit_square([0, 1, 2], [0, 1])
@test_throws ArgumentError Lighthouse.area_under_curve_unit_square([0, 1, 2], [0, 1])
@test ismissing(Lighthouse.area_under_curve_unit_square([], []))
@test isapprox(Lighthouse.area_under_curve_unit_square(collect(0:0.01:1),
collect(0:0.01:1)), 0.5;
atol=0.01)
collect(0:0.01:1)), 0.5;
atol = 0.01)
@test isapprox(Lighthouse.area_under_curve_unit_square(collect(0:0.01:(2π)),
sin.(0:0.01:(2π))), 0.459;
atol=0.01)
sin.(0:0.01:(2π))), 0.459;
atol = 0.01)
end